Hybrid Genome Assembly

  • Hybrid genome assembly is an approach that combines sequencing data from two or more technologies to reconstruct genomic sequences with improved accuracy, completeness, and continuity. It most commonly combines short-read sequencing data, such as Illumina reads, with long-read sequencing data generated by technologies such as PacBio or Oxford Nanopore. Short reads generally provide highly accurate nucleotide information, while long reads can span repetitive regions, connect distant genomic segments, and resolve complex genome structures. By combining these complementary strengths, hybrid assembly can overcome several limitations associated with using either short-read or long-read sequencing alone.
  • The main objective of hybrid genome assembly is to produce a genome sequence that is both highly accurate and sufficiently continuous to represent the underlying biological genome. Short-read assembly can provide accurate sequences but may generate many fragmented contigs because short reads cannot resolve long repetitive regions. Long-read assembly can generate longer contigs and improve structural continuity, but the accuracy of the resulting assembly depends on the sequencing technology, basecalling quality, and consensus correction methods. Hybrid approaches use information from multiple sequencing platforms to improve the final assembly.
  • Hybrid assembly is particularly useful for microbial genomes, plant and animal genomes, complex genomic regions, plasmids, and metagenomic research. It is widely applied in bacterial genome reconstruction, pathogen characterization, antimicrobial resistance research, comparative genomics, genome annotation, and evolutionary studies. The appropriate hybrid strategy depends on genome size, complexity, sequencing coverage, read length, sequencing accuracy, and the biological objectives of the study.
  • The most common hybrid assembly strategy combines Illumina short reads with long reads from PacBio or Oxford Nanopore. Illumina sequencing generally produces paired-end reads, commonly around 100–300 bases in length. These reads are typically characterized by high per-base accuracy and relatively low sequencing error rates. PacBio and Oxford Nanopore technologies produce much longer reads, which may span thousands or tens of thousands of bases and, under suitable conditions, considerably longer sequences. Their longer read lengths help resolve repeats, structural variations, and genomic regions that cannot be reliably reconstructed from short reads alone.
  • PacBio sequencing provides long reads through single-molecule real-time sequencing. Its HiFi technology generates highly accurate consensus reads by sequencing circular DNA templates multiple times and combining the observations. HiFi reads provide a useful combination of length and accuracy, making them suitable for de novo assembly, structural variant analysis, and haplotype-resolved genome reconstruction. Oxford Nanopore sequencing generates reads by measuring electrical signals as DNA molecules pass through nanopores. It can produce long and ultra-long reads, which are particularly useful for spanning repetitive regions and connecting genomic segments. The characteristics and error profiles of these technologies differ, so hybrid assembly workflows must account for the actual read type rather than treating all long reads as equivalent.
  • Hybrid assembly can be performed using several strategies. In one approach, long reads are used to construct an initial genome assembly, and short reads are subsequently mapped to the assembly to correct nucleotide-level errors. This strategy is often called long-read assembly followed by short-read polishing. It is particularly useful when long reads provide sufficient continuity but their consensus sequence requires additional correction. The short reads help identify likely substitutions and small insertion or deletion errors, although polishing is limited by the accuracy and coverage of the short-read data.
  • A second approach uses both short and long reads during the assembly process. Hybrid assemblers incorporate information from the two sequencing technologies when constructing genomic sequences. Depending on the software, short reads may help resolve local sequence differences, correct errors, or support graph construction, while long reads provide connections across repeated and complex regions. This strategy can produce a more integrated reconstruction than simply assembling the two datasets independently.
  • A third strategy involves generating separate short-read and long-read assemblies and then comparing or combining their results. The assemblies can be evaluated for completeness, contiguity, accuracy, and structural consistency. However, independently generated assemblies should not simply be concatenated or merged without examining their relationships. Differences may reflect sequencing errors, heterozygosity, repeat resolution, alternative haplotypes, contamination, or genuine genomic variation. Assembly reconciliation therefore requires appropriate algorithms and biological validation.
  • The first stage of hybrid genome assembly is experimental design and sequencing data preparation. Researchers select sequencing technologies according to genome size, complexity, repeat content, expected heterozygosity, available resources, and the desired assembly quality. They then prepare suitable sequencing libraries and generate short-read and long-read datasets from the same biological sample. Using data from the same sample is important because differences between individuals, strains, or tissues can otherwise be mistaken for sequencing or assembly discrepancies.
  • The sequencing data are generally stored in platform-specific formats. Illumina short reads are commonly provided in compressed FASTQ files, while PacBio data may be supplied as BAM or FASTQ files depending on the processing stage. Oxford Nanopore data may be stored as FASTQ files after basecalling, with raw signal information retained in POD5 or, in older workflows, FAST5 files. Researchers should preserve the original files and document the sequencing platform, chemistry, library preparation, read length, basecalling software, and other relevant metadata.
  • Quality control is essential before combining the datasets. Short-read quality assessment commonly examines per-base quality scores, adapter contamination, GC content, ambiguous bases, and duplication levels. Tools such as FastQC, MultiQC, fastp, and Cutadapt are frequently used for this purpose. Long-read quality assessment examines read length distributions, total sequencing yield, quality scores, and possible contamination. Tools such as NanoPlot and NanoStat are commonly used for Oxford Nanopore data, while PacBio workflows use appropriate SMRT Link and related quality-control utilities.
  • An example of preprocessing paired-end Illumina reads using fastp is:

fastp \

    -i reads_R1.fastq.gz \

    -I reads_R2.fastq.gz \

    -o clean_R1.fastq.gz \

    -O clean_R2.fastq.gz \

    –detect_adapter_for_pe \

    –qualified_quality_phred 20 \

    –length_required 50 \

    –html fastp_report.html \

    –json fastp_report.json

  • This command removes detected adapters and filters low-quality or short reads according to the specified settings. The thresholds are illustrative and should be selected according to the sequencing data and assembly requirements. Long-read preprocessing should be performed separately using tools and settings appropriate for the platform. Researchers should avoid applying short-read trimming procedures indiscriminately to long reads because the error profiles and useful sequence characteristics differ.
  • Sequencing coverage is another important factor in hybrid assembly. Coverage represents the average number of times a genomic position is represented by sequencing reads. It can be estimated using the following expression:

[
C = \frac{N \times L}{G}
]

  • where (C) is the approximate coverage, (N) is the number of reads, (L) is the average read length, and (G) is the estimated genome size. For paired-end short reads, the total number of bases from both read sets should be included. For example, 20 million paired-end read pairs with 150 bases per read produce approximately 6 billion bases. For a genome of 100 million bases, this corresponds to an estimated short-read coverage of 60-fold.
  • Coverage should be calculated separately for each sequencing technology because short-read and long-read datasets contribute different amounts of sequence information. A hybrid project may have high short-read coverage but insufficient long-read coverage to span important repetitive regions. Conversely, a long-read dataset may provide good continuity but lack enough accurate short-read coverage for reliable polishing. Coverage requirements depend on genome complexity, read accuracy, ploidy, heterozygosity, and the assembly software. Average coverage also does not guarantee uniform representation across the genome.
  • One widely used strategy is to assemble the long reads first and then polish the assembly using short reads. Long-read assemblers such as Flye, Canu, and Hifiasm can generate an initial assembly, depending on the input technology and read characteristics. Short reads are then aligned to the assembly using tools such as BWA-MEM or Bowtie2. Consensus correction tools can use the alignments to identify and correct supported nucleotide differences. Depending on the workflow, polishing may involve tools such as Pilon or Polypolish. The choice of polishing software should reflect the sequencing data, alignment characteristics, and current tool recommendations.
  • For example, a long-read assembly can first be generated using Flye with Oxford Nanopore raw reads:

flye \

    –nano-raw long_reads.fastq.gz \

    –genome-size 100m \

    –out-dir flye_output \

    –threads 8

  • The –nano-raw option is intended for the relevant raw Oxford Nanopore read type supported by the installed Flye version. Other read types, including higher-accuracy nanopore reads and PacBio HiFi reads, require the corresponding input mode. The genome-size value should be adjusted to the organism being studied.
  • The resulting assembly can then be indexed and aligned against cleaned Illumina reads using BWA:

bwa index flye_output/assembly.fasta

bwa mem \

    flye_output/assembly.fasta \

    clean_R1.fastq.gz \

    clean_R2.fastq.gz \

    > short_reads.sam

SAMtools can convert and sort the alignment file:

samtools view \

    -bS short_reads.sam \

    -o short_reads.bam

samtools sort \

    short_reads.bam \

    -o short_reads.sorted.bam

samtools index short_reads.sorted.bam

  • Pilon can use the assembly and short-read alignments to identify supported corrections. A simplified example is:

pilon \

    –genome flye_output/assembly.fasta \

    –frags short_reads.sorted.bam \

    –output polished_assembly \

    –outdir pilon_output

  • Pilon was designed for polishing assemblies using suitable short-read alignment data, and its use should follow the requirements of the installed version. The alignment file must be properly prepared, and the assembly should be evaluated after polishing. Repeated polishing rounds should not be applied automatically because corrections can sometimes introduce errors, particularly in repetitive regions or where short-read mapping is ambiguous.
  • Another important approach is direct hybrid assembly using software designed to incorporate short and long reads together. Unicycler is widely used for bacterial genome assembly and can combine Illumina short reads with long reads from compatible sequencing technologies. It uses short-read information to support accurate sequence reconstruction and long-read information to resolve connections between genomic regions. It is particularly useful for bacterial chromosomes and plasmids, although its suitability depends on the organism, input data, and workflow.
  • An illustrative Unicycler command is:

unicycler \

    -1 clean_R1.fastq.gz \

    -2 clean_R2.fastq.gz \

    -l long_reads.fastq.gz \

    -o unicycler_output \

    -t 8

  • Here, -1 and -2 specify the paired-end short reads, -l specifies the long reads, -o defines the output directory, and -t sets the number of threads. The long-read input must be in a format and quality range supported by the installed Unicycler version. Unicycler is primarily intended for bacterial genome assembly and should not be assumed to be a general-purpose solution for large eukaryotic genomes.
  • Other tools can support hybrid assembly and polishing workflows. MaSuRCA combines short- and long-read information in its assembly process and has been used for different genome types. SPAdes includes hybrid assembly capabilities through its hybridSPAdes mode for suitable datasets. Flye and Canu are commonly used for long-read assembly, after which short-read polishing can be performed with suitable correction tools. Hifiasm is designed primarily for PacBio HiFi assembly, and its output can be further evaluated or supported with complementary data where appropriate. The best software choice depends on whether the project involves a microbial isolate, a metagenome, a diploid animal, a plant genome, or another biological system.
  • Hybrid assembly is particularly valuable for bacterial genome reconstruction. Many bacterial genomes are relatively small, but they can contain repeated sequences, insertion sequences, plasmids, genomic islands, and mobile genetic elements that complicate short-read assembly. Long reads can span these regions and help connect genomic segments, while short reads can improve nucleotide-level accuracy. Hybrid approaches can therefore support the reconstruction of more complete bacterial chromosomes and plasmids, which is useful for identifying antimicrobial resistance genes, virulence factors, and genomic differences among strains.
  • Plasmid reconstruction is an important application of hybrid assembly. Plasmids are often circular DNA molecules that may contain repeated elements, mobile genetic elements, and antimicrobial resistance genes. Short-read assemblies may fragment plasmid sequences or fail to distinguish similar plasmids from chromosomal regions. Long reads can span plasmid structures and provide information about their organization, while short reads can help correct sequence errors. However, plasmid reconstruction still requires careful validation because repeated sequences and multiple related plasmids may produce ambiguous assembly paths.
  • In metagenomics, hybrid assembly can combine short and long reads from complex microbial communities. Long reads may connect genes and genomic regions within the same organism, while short reads provide additional sequence information and may support error correction. Hybrid metagenomic assembly can improve the reconstruction of metagenome-assembled genomes and help characterize microbial communities. Nevertheless, strain variation, unequal organism abundance, shared genomic regions, and contamination remain important challenges. The resulting assemblies and genome bins should be evaluated using appropriate completeness, contamination, and taxonomic consistency methods.
  • Hybrid assembly also has applications in plant and animal genomics. Large eukaryotic genomes frequently contain extensive repetitive DNA, duplicated genes, and complex chromosome structures. Long reads can improve contig continuity, while accurate short reads can contribute to sequence validation and correction. Additional technologies, including Hi-C sequencing and optical mapping, may be needed to organize contigs into chromosome-scale assemblies. For highly heterozygous or polyploid genomes, specialized assembly and phasing methods may be required to distinguish homologous chromosomes and different genomic copies.
  • Assembly quality assessment is essential for evaluating hybrid genome assemblies. Researchers should examine total assembly length, number of contigs, largest contig, N50, L50, GC content, and sequencing-read coverage. N50 provides a measure of contiguity, but it does not establish sequence accuracy or biological completeness. A highly contiguous assembly may still contain misassemblies, collapsed repeats, or incorrect structural connections. Quality assessment should therefore combine multiple statistics with independent evidence.
  • QUAST is commonly used to assess assembly statistics and compare assemblies with a reference genome when one is available. A basic example is:

quast.py \

    polished_assembly.fasta \

    -o quast_output

  • A reference genome can be supplied for additional comparisons:

quast.py \

    polished_assembly.fasta \

    -r reference_genome.fasta \

    -o quast_reference_output

  • Reference-based evaluation should be interpreted carefully if the reference is genetically distant from the organism being assembled. Genuine biological differences may be identified as apparent discrepancies, and reference bias may conceal novel or structurally different regions.
  • BUSCO can assess the completeness of conserved single-copy orthologous genes in an assembly. It is useful for estimating gene-space completeness, but it does not independently validate the entire genome or guarantee that all structural connections are correct. The lineage dataset should be selected according to the organism and the datasets available in the installed BUSCO version.
  • An example BUSCO command is:

busco \

    -i polished_assembly.fasta \

    -m genome \

    -l bacteria_odb12 \

    -o busco_output \

    -c 8

  • This example is intended for a bacterial genome. A suitable lineage dataset must be selected for other organisms.
  • Read mapping provides another important validation method. Both short and long reads can be mapped back to the final assembly to evaluate coverage, alignment consistency, and possible assembly problems. BWA-MEM or Bowtie2 can be used for suitable short-read datasets, while minimap2 is widely used for long-read alignment. SAMtools can process alignments and generate mapping statistics. Coverage profiles, discordant paired-end reads, split alignments, and unusual read-depth patterns may provide evidence of possible misassemblies or unresolved genomic structures.
  • For example, long reads can be aligned to a hybrid assembly using minimap2:

minimap2 \

    -ax map-ont \

    polished_assembly.fasta \

    long_reads.fastq.gz \

    | samtools sort -o long_reads_mapped.bam

samtools index long_reads_mapped.bam

samtools flagstat long_reads_mapped.bam

  • The map-ont preset is appropriate for relevant Oxford Nanopore reads. PacBio HiFi reads should use a suitable HiFi-specific preset. The mapping results should be interpreted together with assembly graphs, coverage information, and other independent evidence. High read-mapping rates alone do not prove that an assembly is correct.
  • Hybrid genome assembly has several advantages. It can improve nucleotide accuracy compared with an unpolished assembly generated from error-prone long reads, increase contig continuity compared with short-read-only assembly, and improve the reconstruction of repetitive and structurally complex genomic regions. It may also reduce the number of unresolved gaps and provide more complete representations of chromosomes and plasmids. These benefits can support more reliable gene prediction, genome annotation, comparative genomics, and evolutionary analysis.
  • Despite these advantages, hybrid assembly also has limitations. Generating multiple sequencing datasets increases experimental cost, sample preparation requirements, and data-management complexity. The sequencing technologies may introduce different biases, and differences between datasets can complicate alignment and correction. If short reads are mapped to an incorrect or highly repetitive region, polishing algorithms may introduce false corrections. Long-read coverage may be insufficient to resolve complex repeats, while short-read data may fail to distinguish closely related genomic copies. Hybrid assembly therefore does not automatically guarantee a complete or error-free genome.
  • Another limitation is the potential for sample inconsistency. If short reads and long reads originate from different biological samples, strains, or individuals, genuine genetic differences may be mistaken for sequencing errors. This can lead to incorrect consensus sequences or misleading assembly structures. Whenever possible, both sequencing datasets should originate from the same biological material, and sample identity should be confirmed using suitable quality-control methods.
  • Reproducibility is essential in hybrid genome assembly. Researchers should preserve raw short-read and long-read files, document sample metadata, record sequencing platforms and library preparation methods, and retain relevant raw signal files when available. Software versions, command-line parameters, preprocessing decisions, assembly settings, polishing rounds, and quality-control results should also be recorded. Checksums can be used to verify file integrity, while workflow managers such as Snakemake and Nextflow can organize the analysis into reproducible pipelines. Conda environments and container technologies such as Docker and Apptainer can help preserve software dependencies.
  • The final assembly should be accompanied by appropriate documentation, including the estimated genome size, sequencing coverage for each technology, assembly software and version, polishing procedures, evaluation statistics, and any known limitations. Where appropriate, assemblies and supporting sequencing data can be deposited in public repositories such as the European Nucleotide Archive, GenBank, or the DNA Data Bank of Japan. Human genomic data and other sensitive datasets require appropriate access controls and compliance with relevant ethical and legal requirements.
  • Hybrid genome assembly is an important approach in modern bioinformatics because it combines the complementary advantages of short-read and long-read sequencing. Short reads contribute accurate nucleotide information, while long reads provide the continuity needed to resolve repetitive regions and complex genome structures. Tools such as Unicycler, SPAdes, MaSuRCA, Flye, and Canu support different hybrid assembly strategies, while polishing and validation tools help improve and evaluate the resulting sequences. Successful hybrid assembly depends on suitable sequencing coverage, consistent sample identity, appropriate software selection, and rigorous quality assessment. When carefully designed and validated, hybrid assembly can produce more accurate and complete genome representations for microbial, plant, animal, and other genomic research.
Author:

Leave a Reply

Your email address will not be published. Required fields are marked *